Articles

Biggest Impact: Devices or Magnetics?

Leave a reply

Comments on Rap Session 1 at APEC 2018

The Rap Sessions 1 last year concluded that “we have all the topologies we need” with flyback and LLC converter declared undisputed champions!? The same moderator and several returning panel members who adjudicated that last year changed this year their hats from topology to become either MAGNETICS or DEVICES experts!? With topologies declared dead on arrival, they invented an artificial contest for Rap Session 1 at this APEC 2018 entitled: BIGGEST IMPACT ON POWER CONVERSION: DEVICES OR MAGNETICS? This is like comparing Apples and Oranges! The only thing they have in common is that they are just components forced to operate at switching frequencies 100 times higher than needed!

UTILIZATION OF DEVICES AND FERRITE MATERIALS?

By doing so APEC 2018 conference managed to completely switch cause and effect in Power Electronics. Topology is a cause and components like switching devices, magnetics and capacitors are just mere consequences! They are the last one to be chosen based on the limitations imposed by switching methods and their topologies! Hence, bad topologies (most present ones!) do result mostly in bad use of all components! Now those device and magnetics “experts” wants us to believe otherwise that new devices and magnetics materials will fix any bad topology!? At APEC 2018 at Rap Session 1 they could not agree only on one thing: how high switching frequency must be, 2MHz or 20MHz! In fact, it was device manufacturers, first as power management companies for the past 50 years, and now the new kid on the block, GaN device manufacturers, who imposed 2MHz on ferrite manufacturers and now demand from them 10MHz ferrite material!? Note that the discussion in this first part is strictly referring to the use of ferrite materials for transformers and not for inductors. Inductors are covered in later Section with new Coreless Step-down Converter Topology.

Discussions at Rap Session:

  1. We have all the ferrite we need… Magnetics 2MHz experts say!?
  2. We need new ferrites for 10MHz operation…Device experts say!

Yet, both Device and Magnetics experts do not recognize each other’s fundamental flaws and how their exclusive focus on 10 MHz switching is completely misguided as further analyzed below.

HIGH FREQUENCY WARNING IN 1988

I have warned Power Electronics community 30 years ago against such “Safe at 10MHz switching speed” back in 1988. The attached cover of 1988 Power Technics magazine shows me running at speed of 500kHz with Integrated Magnetics Isolated Cuk converter topology and its prototype. What follows is another prototype based on Hybrid Switching Method and corresponding Isolated Topology operating at 50kHz (note the reduced running speed!). This is my second warning after 30 years.

While everyone went to 10 times higher switching frequencies of 5MHz I came down to 10 times lower switching frequency of 50kHz and my reduced running speed on attached revised version of the same drawing! Only 100-time difference but with bonus record 98% efficiency and still reduced size!

QUESTION TO DEVICE EXPERTS BY MAGNETICS EXPERTS

  1. Utilization of devices?Have you ever seen Magnetic expert question Device expert for its 1% device utilization at any switching frequency let alone 10MHz?What is Device’s current, voltage and power ratings in actual topologies relative to their maximum capability! Please see enclosed drawing of buck converter used for 48V to 1V, 100A having 1% power utilization of its synchronous rectifier GaN device, or an “overkill” factor of 100!
  2. Hard switching at 10MHz?Did Magnetics expert ever ask Device experts why they are using hard switching at 10MHz (full square wave current of 100 A at turn-on and turn-off of the switch) instead of soft switching used even at 50kHz in proper topologies!The device switching performance is likewise 100% dependent on topology and its ability to operate them at Zero Voltage (ZV), Zero Current (ZC) or both at both switching instances!Why do device experts ignore topologies which are turning on and turning off synchronous rectifier switch at zero voltage and zero current? Then the switch does not even “know” it is switching since at critical switching instances it “sees” only zero current and zero voltage!

QUESTIONS TO MAGNETICS EXPERTS BY DEVICE EXPERTS

  1. Ferrite material utilization?Did Device experts ever ask Magnetics experts what their ferrite material utilization is (ratio of actual transformer flux density relative to its maximum flux capability)? Why is it only 0.1% for 2MHz optimized ferrite materials, or an “overkill” factor of 1000 to 1? Please see enclosed drawing of the ferrite properties of the PC200 material from TDK.
  2. Size reduction with frequency increase?Any size reduction with 40 times increased switching frequency?Why device experts never asked magnetics experts why the size of transformer is staying the same despite the 40 times reduction of frequency from 2MHz to 50kHz?While DEVICE and MAGNETICS experts are praising each other for their advanced 2MHz switching operation, let us turn to a lowly expert on Switching Methods and Topologies.

SWITCHING METHODS AND TOPOLOGIES EXPERT OPINION!

The above answers can only come from Topology experts which were sidelined last year. A more appropriate question to ask now is:

What Comes First: Switching Methods or Topologies?

It is an absence of the appropriate new optimum Switching Methods and their resulting consequence, various converter Topologies, which has stalled Power Electronics for last 70 years! We are now for the first time able to provide simultaneously optimum utilization of switching devices and best utilization of magnetic components with new Switching Methods, Resonance Scaling Method and related novel Topologies. The best yet, we DO NOT NEED 2MHz switching for size reduction as 100kHz works just fine!

MAGNETICS FOR INDUCTORS

How about ferrite cores used for inductors in converters like buck, boost and flyback. It was projected that 50MHz switching frequency would be needed to eliminate ferrite cores (see article in powerelectronics.com) This would not even reduce the magnetics size as the same cross section of the coil winding would be needed!? Yet, with right Switching Method and Topology, the ferrite core is not only eliminated at 50kHz but even reduced to short copper traces on PCB eliminating the need for windings.

CORELESS STEP-DOWN CONVERTER ELIMINATES CORES AND WINDINGS

One such converter is shown using new Storageless Switching Method and related novel Resonance Scaling Method. This drawing compares the buck converter with Coreless step-down DC-DC topology which eliminates magnetic cores and windings even at 50kHz but operates with duty ratio control and has one cycle transient response. The link to detailed powerlectronics.com article is at goo.gl/3N7qBp

BUCK CONVERTER RETIREMENT

While buck converter is used for last 70 years, now the times has come for its well-deserved retirement! Shown below is comparison buck converter with New Coreless Step-Down DC-DC. Both are regulating output DC voltage with simple PWM duty ratio modulation but with these crucial differences:

BUCK CONVERTER

  1. Large ferrite core even at 20A DC load current.
  2. Inferior transient response due to energy storage.
  3. Eight buck modules needed in parallel to fix transient response.
  4. Used exclusively to provide 12V to 2V, 100A eight module VRD.
  5. Ten-billion-dollar industry relies exclusively on this “solution”!

CORELESS STEP-DOWN CONVERTER

  1. Single IC chip with 4 GaN devices
  2. One Cycle transient with single module.
  3. No magnetics core at 50kHz.
  4. No coil windings at 50kHz.
  5. Two short copper traces of few mm each for two inductors.
  6. Ultra-high efficiency of 99%.
  7. No heat sinks needed.
  8. First Power Supply on a Chip (PwrC) for 12V to 2V, 100A.

There are number of extensions including galvanic isolation with much reduced transformer size at 50kHz switching!

REAL ADVANTAGES OF GaN DEVICES

To dispel the notion that I am advocating to go from Transistors to Vacuum tubes, there is a great news for both GaN devices and ferrite materials when they are optimized for 50 kHz to 100kHz operation. Terrific low loss ferrite materials already exist with near 100% utilization. GaN manufacturers, however completely missed this with their ultra-high frequency direction! Yet, GaN devices true advantages are:

  1. Planar technology enabling all devices and their interconnections can be implemented on a single chip.
  2. Integrated drive circuitry on the same chip with power devices.
  3. Making power switches and drive circuitry into an Integrated Power Circuit (IPC) for power processing like Integrated Circuits (IC) did for signal processing electronics.
  4. Paralleling number of devices for much reduced conduction losses.
  5. Scaling up to higher current and power taking advantage of five-times smaller size of devices compared to MOSFETs.

It does not take much to see that these GaN advantages will NOT be realized at 10MHz switching but will be much easier realized at 100 kHz switching with much improved efficiency and reduced size to boot!

Those GaN manufacturers who recognize first the need to implement and optimize their devices with new System Technology will not only be well rewarded but will also make critical contribution to advancement of Power Electronics field.

CONCLUSION

Buck converter is bad even for a single module let alone for 8 modules in parallel! Thus, it should be abandoned in favor of modern solutions. This is just one of a number of “little box solutions!” which came owing to implementation of new Switching Methods and Resonance Scaling Method resulting in state-of-the-art novel Converter Topologies. The most important is that several solutions include also isolated converters with transformer size at 100kHz even smaller than 2 MHz present designs…Oh, and one more thing: your comments are very welcome!

In addition, those who want to collaborate on this path to Power Electronics for 22nd Century NOW feel free to contact me via email on cuk@teslaco.com.

Dr. Slobodan Cuk, formerly Professor at Caltech.

GaN EXPERT QUOTES

“Magnetics are catching up with higher frequencies possible with GaN.”

“GaN is not 10 times ahead of signal processing ICs. The problem is that silicon-based controller is not designed to operate at higher than 2-3 MHz. They could be, but the controller folks are lagging the GaN folks.”

“Yes, finally! There is also a need for higher speed digital control ICs for DC-DC conversion. GaN devices can be efficiently hard-switched at 10 MHz, but digital controller is usually limited to ~1 MHz.”

Comments on Slobodan Cuk’s article:

Michael Miller – President at Miller International Group

“As a hard magnet guy, this is a comparatively easy to follow explanation/review of a very complex issue. Is Ferrite the only magnetic material used in these devices? Or are there hybrids? Thanks Slobodan.”

Slobodan Cuk – CEO at TESLAco

“Michael, Yes, ferrite materials are the only available ferromagnetic materials with acceptable core losses per unit volume for frequencies above 20kHz. If you refer to the hybrids as alloys of various combinations of Nickel, Iron and Cobalt, there are none so far to my knowledge.

TO ALL, HERE IS A LINK TO MUCH ENLARGED AND IMPROVED VERSION NAMED DEVICES OR MAGNETICS? EXT VER https://goo.gl/nW5zrr . I added video clips, additional drawings and comments. Shown below is Excel spreadsheet derived from full TDK specifications for PC 200 material which highlights the key parameters responsible for determination of the size. It shows clearly that 30 times increase in switching frequency from 100kHz to 3MHz does NOT reduce size at all! All it does is reduce efficiency dramatically due to skin and proximity effects at high frequency and increased AC resistance of copper windings. Obviously, this results in cost increase as well opposite of claimed cost reduction due to bogus reduction of transformer size. NIKOLA TESLA was the first one who showed dramatically the skin effect at high frequency and how the high frequency currents crowd to the surface of conductors dramatically increasing AC resitance over DC resistance of windings. He connected himself in a circuit with his high frequency generators and was melting copper bars without any harm to his body. In fact, this is now one of the methods used for therapeutic purposes!

Coupled-Inductor SEPIC converter

Leave a reply

Perfectly Coupled-Inductor SEPIC converter
Reduces to WHICH Converter?

Analysis of Equivalent Circuit Transformations
by Professor Slobodan Ćuk

Perfectly Coupled-Inductor SEPIC Converter

Abstract

For quite a while now, the various “shortcuts” were circulated on LinkedIn media, such as IEEE Power Electronics Society and various websites extolling the virtues of the PWM switch small-signal model! In particular, the claim is made that the “magic shortcuts” could reduce frequency response of a Perfectly Coupled-Inductor Nonisolated SEPIC converter (CI Converter from now on!) to that of a Boost converter as included in a prior discussion to this group!?
However, the detailed analysis bellow reveals:

  1. The PWM small-signal switch model is not even needed! The sequence of correct equivalent circuit transformations would, even without any small-signal model, reduce CI Converter to ordinary nonisolated polarity inverting buck-boost converter! The addition of small-signal model only contributes to hide transformation mistakes.
  2. It then follows, even without small-signal model that CI Converter frequency response is that of buck-boost converter and not boost with its right-half-plane zero! No magic there either!
  3. This analysis points out to the critical incorrect circuit transformations leading to this erroneous result.
  4. It also points out at other incorrect simplifying assumptions, which led to boost converter circuit!
  5. As invalid transformations and simplifications are correctly identified here the correct final result of the nonisolated flyback converter is obtained!

Checking for contradictions of final result

This analysis also points out that when one obtains “surprising” result like boost frequency response, one should thoroughly scrutinize for obvious contradictions! In this case, it is simply impossible that the final result could have small signal frequency response of the boost converter and DC conversion gain of an entirely different buck-boost converter as claimed by PWM switch model!?

Detailed Analysis of CI Sepic converter

This analysis will also introduce truly “magical shortcut”, which can be used to determine what switching converters qualify for implementation of the Coupled-Inductor and Integrated Magnetics Methods (see 1979 patent posted to this group)!

Fig. 1 Basic SEPIC converter and its AC circuit

Test to verify Coupled-Inductor applicability?

First, we establish the fact that the qualifying converters must have at least two inductors before coupling. Coupling of Inductors does not change DC voltage gain which existed before the coupling. Instead, coupling only affects the distribution of the AC ripple currents between two winding as I discussed in numerous publications and in my patent on Ćuk converter and Coupled-Inductor SEPIC converter (see thread posted before!)

What makes given switching converter converter eligible for implementation of Coupled-Inductor Method I introduced over 44 years ago!

Fig.1 has original SEPIC converter! By shorting capacitors and DC voltage source, we obtain bottom AC circuit model. This is a true magic! Note that the two inductors are connected in parallel in AC model! Note also that the two switches, ideal switch S and diode CR, will generate identical common rectangular-wave voltage drive for any duty ratio D. This then qualifies both inductors to be placed on a common core in a transformer-like configuration. Note also that there could be only one turns ratio that of 1:1. Arbitrary turns ratio is not allowed, unlike in real AC transformer. Another difference is that the DC currents of each separate inductor are added together as dictated by the dot connections!

The two parallel inductors could obviously be replaced with a single common inductor L in case of a perfectly coupled winding with no leakage, when L1=L2 = L. The trick is now to determine the correct way of implementing such coupling in original DC-DC SEPIC converter. However, we first introduce the two incorrect transformations!

Two incorrect transformations

Fig. 2 further explains that point. Common inductance can be shown in circuit model as either on primary side (top circuit) or on secondary side (bottom side) of the ideal 1:1 isolation transformer! Note that marked termination points 1 & 2 for L1 and 3 & 4 for L2 determine also correct dot connection of the isolation transformer. Hence two possible circuits could be contemplated as in Fig 2 below.

Fig. 2 Two incorrect transformations with floating transformer (no common ground)

Model in which common inductance was on a secondary side (bottom drawing) is obviously wrong, since the input voltage source is NOT connected to converter at all!) On would then think that model in Fig. 2 (top) in which common inductance is retained on primary side would be OK, since now input source is connected and one would have big sigh a relief. This is, however, also wrong! Both transformations have the same problem! Both inductors in original topology are floating inductors and cannot be directly replaced with either of two inductor circuit positions.

It is interesting to note that the converter of Fig. 2 (top) is does not have steady-state and hence it does not even exist! The capacitor Cc is charging during OFF- time of switch but lacks discharge path!?

Wrong boost equivalent circuit model?

It is interesting how a desire to further simplify the wrong model of Fig.2 (top) leads to wrong boost converter model. The dubious argument is made that capacitor Cc for some reason (!?) does not participate in power conversion (although it did in SEPIC converter!) and can be therefore eliminated by shorting it to result in boost converter or Fig. 3!?

Fig. 3 Wrong boost converter model using false assumption that capacitor Cc can be SHORTED?!

This is PWM switch has led to incorrect conclusion that the Perfectly Coupled SEPIC converter using PWM switch model will have-boost frequency response!?

Correct Transformation

However, the input inductor can be relocated to the bottom leg so that terminal 1 and terminal 4 are common forming a 3-port network with a common terminal 1 & 4. Now, both transformations discussed above have the same identical correct solution obtained by shorting terminals 2 and 4 as in Fig. 4.

Fig. 4 Relocation of the input inductor to bottom leg and shorting the other two terminals 2 and 3.

This ultimately placies voltage source across coupling capacitor Cc and results in buck-boost final equivalent circuit. Note that capacitor Cc is now placed directly across ideal voltage source. This is a real reason why coupling capacitor Cc disappears in the correct buck equivalent buck-boost converter. The fact is, that in the case of the nonideal voltage source this capacitor would remain in the model and would result in buck-boost converter with inclusion of capacitor Cc and third-order dynamic model!

Final correct buck-boost equivalent circuit model

Fig. 5 Final correct buck-boost equivalent circuit.

Conclusion

Number of incorrect transformations and incorrect simplifications like shorting capacitor Cc led to wrong boost small-signal model!? Moreover, after deriving incorrectly boost small-signal frequency response, Vorperian continues to prove correctly that the DC voltage conversion is that of buck-boost!?

This would be the very first converter ever, which has frequency response of boost converter and DC voltage gain of buck-boost converter! That alone should have given the red flag to Vorperian and Ridley that something went amiss in their “magic” shortcuts!

Note regarding PWM switch model!

PWM switch model starts, as in this example, with small-signal model first, while steady-state DC model is always an after-thought!?

Yet, it is steady-state model which defines equilibrium around which small signal perturbations are made and should be the starting point!

State-Space Averaging Model

This is the case with State-Space Averaging Method, which determines first correct general steady-state (DC) model for any PWM switching converter, for those known 50 years ago as well as those invented ever since! It does not depend on “shortcut tricks” to massage the switch into an artificial 3-terminal double pole single-throw switch!

What about isolated converters or even 3-switch converters with isolation like forward and flyback converter with voltage clamp? Note that 3-switches cannot be manipulated into a single-pole double-throw ideal switch. The isolation in flyback and forward simply makes any such PWM switch transformation impossible and/or incorrect!?

Think how you would find correct frequency response of the Coupled-Inductor Isolated Sepic converter or even Integrated Magnetic SEPIC converter. With PWM switch model it is impossible. With State-Space Averaging Method this becomes a trivial extension by describing correctly mutual coupling inductances in matrix formulation.

Myth of the Moving Time Average!

Ever since my introduction of State-Space Averaging method in 1976 PhD thesis and 1976 IEEE conference paper, many revision attempts were made to explain averaging circuit method. They all followed original justification for boost small signal model in Gene Wester PhD thesis invoking “moving time average! It was claimed ever since that this somehow eliminates the ripple currents and voltages and results in smooth ripple-free continuous waveforms.

This continues even to the present day even though it has no scientific basis whatsoever! Another related argument is that just replacing PWM switch with dynamic model circuit gives you better “circuit insight” equal lacks any scientific basis!

It appears that my advisor and mentor late Prof. David Middlebrook was the only one who correctly understood my State-Space Averaging Method! In the preface of our 1981 first 2 volume paperback edition of Power Electronics books Prof. Middlebrook lamented:

“… If the models for all such converters are the same, it should be possible to derive this unique model without having to specify in advance any particular converter. This problem was solved in a very elegant manner by Slobodan Ćuk. In his 1976 PhD thesis he introduced the analysis method of State-Space Averaging which in a single stroke eliminates the switching process from consideration and exposes the desired dynamic response. From this model came the same unique small signal equivalent circuit model, which is now called the canonical model. Again, with the clarity of insight, the form of the model becomes “obvious”. It contains the three essential properties of any DC to DC converters, namely DC conversion, low pass filter and conversion ratio adjustment by a control signal.”

Extension of SSA to Hybrid Switching Method

State-Space Averaging is naturally extended now to special kind of resonant switching I introduced and called Hybrid Switching Method! This method has all the benefits of resonance but using duty ratio PWM control. It does also have 3 switches (one transistor on primary and two diodes on secondary) which automatically disqualifies PWM switch model! In addition, the resonance is essential as it fundamentally changes DC conversion ratio! PWM switch model as its names says is limited to PWM switching converters with no resonance and with special limiting single-pole double throw switch with a common terminal!

This SSA extension is not suitable for classical true resonant and quasi-resonant switching converters and latest LLC converter topologies, in which resonance extends across both switching subintervals. This is just fine, as these resonance methods are in many ways deficient to Hybrid Switching Method which contains the resonance in only one subinterval!

Conclusion

The PWM switch model and other preferable” shortcuts” are just a myth propagated by engineers who do not want to invest time to learn and expect that SPICE model of such simplified circuit model (albeit wrong) will give them the fast and correct results. Recent 2018 article on modelling says: “Compared with State-space averaging model, the PWM switch model is simpler and circuit oriented with physical insight.” No one ever explained what those better circuit and physical insights are?!

Another myth is that State-Space Averaging Method is very complicated and should not be used. The truth is that this method will give you correct analytical results for both steady-state and dynamic model for all converters of practical interests!

Would you ever want to have perfect coupling?

The answer is no, never! The question remains to be explained why this is the case if the above analysis makes the case for simpler 2nd order dynamic model of Perfectly Coupled SEPIC converter!?

What Comes First: Devices or Topology

Topologies Comparison

Topologies Comparison

Present Power Electronics Systems are built upon and resemble an INVERTED pyramid which is by definition unsustainable and ready to collapse at any moment. Why? At its bottom it has the buck converter as its foundation which than supports the next level single ended isolated converter like flyback and forward to culminate on the third level into bridge type converter with even more switching devices. Consequence: buck converter problems lead to collapse of the whole structure irrespective what switching devices are used.

The buck converter does have huge problems: device voltage stresses much higher than output regulated DC voltage and inherent hard switching operation. Moreover, the use of the inductor for filtering output ripple currents leads to inherent problems in scaling up the power due to the requirement to pass DC load current so that even modest DC load current and resulting air-gap just a fraction of millimeter leads to complete waste of the magnetic material and ineffective filtering. The new PWM/resonant switching method results in converter topologies which eliminate output inductor with DC bias and yet offering much more effective ripple current filtering.

The new switching methods and resulting converter topologies result in much reduce device voltage stresses. In the isolated converters this translated into reduction of transformer fluxes by an order of magnitude and consequent huge size reduction and efficiency improvements. The next example demonstrates that the optimum converter topology is of the primary importance for system performance and that actual switching devices utilized have only secondary importance in system improvements.

Converter Graphic

Converter Graphic

Comparison of the two switch buck converter with the three switch Cuk-buck2 demonstrates the huge differences in device stresses for 48V to 1V step down conversion. The synchronous rectifier diode in buck converter has 100V voltage stress versus 2V voltage stress in Cuk-buck converter. Hence a 10kW rated switch must be used instead of the 200W rated switch! Moreover, the diode is turned ON and turned- OFF at full 100A current in buck converter. In Cuk-buck2 the diode is turned both ON and OFF at zero current. Therefore, it does not even know that it is switching although it transfers huge current to the load. The duty ratio in buck converter is extremely small at 0.02 in order to make a 50V to 1V step-down buck converter. The Cuk-buck2 operates at 50% duty ratio and results in reduction of magnetics size by at least 50%.

Large Voltage Graphic

Large Voltage Graphic

In addition to elimination of switching losses Cuk-buk2 also has an isolated extension which retains all the above advantages, while the isolated version of the buck converter, the forward converter, suffer from the same disadvantages as the buck converter.

Device performance depends 100% on topology and not the other way around!